Solving harmonics elimination problem in three-phase voltage-controlled inverter using artificial neural networks

Abstract EN

A novel concept of application of Artificial Neural Networks (ANN) for generating the optimum switching functions for the voltage and harmonic control of DC-to-AC bridge inverters is presented.

In many research, the neural network is trained off line using the desired switching angles given by the classic harmonic elimination strategy to any value of the modulation index.

This limits the utilisability and the precision in other modulation index values.

In order to avoid this problem, a new training algorithm is developed without using the desired switching angles but it uses the desired solution of the elimination harmonic equation, i.e.

first harmonics are equal to zero.

Theoretical analysis of the proposed solving algorithm with neural networks is provided, and simulation results are given to show the high performance and technical advantages of the developed modulator.